Editing and revising text is tedious and error-prone.
Word processors have hardly any support for working with words.
Most functions are character-based.
Writers must translate high-level goals into low-level editor functions.
Writers have to adapt their writing strategies.
This results in typical errors: missing or duplicative words, lost words, extra words, s/v agreement problems.
Even conceptually simple tasks require many steps.
Hmm. This seems to be on computer science. It takes eight steps to switch “revising and editing” to become “editing and revising.” Let’s see:
Okay, I might have missed something that would make it just as bad in real-time. Of course, I don’t think that is too difficult.
Back to the speaker
Norman (1981): Widely accepted classification of errors; editing errors can be interpreted as action slips.
Norman (1983): Many slips result from bad design and are preventable by more appropriate design.
How could the design be improved?
Hypothesis: Tools and functions operating on the same level on which writers think and talk bout texts makes editing easier.
Language-aware editing functions
power tools, not checkers (You still have to cut the wood to build the table, even with a power saw.)
functions need linguistic knowledge
useful functions can be implemented even with small linguistic resources
Writing research + computational tools
Prototypical implementation of selected functions
Target group: experienced writers
target language: German
no development of a new editor: XEmacs as test bed
inspired by source code editors for programmers (not just stream of characters)
Fascinating stuff and far over my head.
I did actually understand what was going on by the end. Computational linguistics and composition… Words instead of letter. Interesting.